|Publication number||US5365429 A|
|Application number||US 08/003,071|
|Publication date||Nov 15, 1994|
|Filing date||Jan 11, 1993|
|Priority date||Jan 11, 1993|
|Also published as||DE69425594D1, DE69425594T2, EP0609922A2, EP0609922A3, EP0609922B1|
|Publication number||003071, 08003071, US 5365429 A, US 5365429A, US-A-5365429, US5365429 A, US5365429A|
|Inventors||Charles S. Carman|
|Original Assignee||North American Philips Corporation|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (9), Non-Patent Citations (40), Referenced by (61), Classifications (9), Legal Events (5)|
|External Links: USPTO, USPTO Assignment, Espacenet|
1. Field of the Invention
The present invention relates generally to methods for aiding in the detection of cancer by utilizing computer analysis of radiologic images to identify spots corresponding to objects associated with malignancy. In its more particular respects, it relates to methods of aiding in the detection of breast cancer by computer identification of microcalcifications and clusters thereof in mammographic images.
2. Description of the Related Art
Breast cancer is one of the primary causes of death for women in western societies. Because the vast majority of deaths due to cancer that originated in the breast could be prevented by early detection, several national health organizations in the United States now recommend that all women over the age of 40 have regular screening mammograms. In fact, the National Cancer Institute has set a goal of 80% compliance by the end of the decade. To reach this goal, the number of mammograms taken and read in the U.S. would have to double, and in some localities, more than triple. An increase of this magnitude would overload the current capacity to take and interpret mammograms.
The reading or interpretation of screening mammograms is an art that requires extensive experience and attentiveness to detail. While the radiologist's primary sign for cancer is a mass visible on the mammogram, one of the more sensitive signs is the presence of small relatively bright spots (in film-screen mammography) corresponding to locally increased X-ray attenuation due to minute (<5 mm in maximum dimension in the film image) deposits of calcium salts known as microcalcifications, each generally of irregular shape, and which are arranged in clusters. In fact, clustered microcalcifications are often the only sign indicating an early in situ malignancy. Unfortunately, these small spots generally appear in the images as obscured by gradations in background produced by surrounding tissue and their visibility is limited by the image's resolution, contrast and signal to noise ratio. Consequently, the probability of their detection by an experienced radiologist is not as high as desirable, making double independent readings a common practice to achieve acceptable results.
A method of computer identification of microcalcifications in digitized mammogram images is known from U.S. Pat. No. 4,907,156, which method is also applicable to a related problem of detection of spots in digitized chest X-ray images corresponding to lung nodules. Therein, spatial filtering was employed to deemphasize both low and very high spatial frequencies in the image attempting to increase the conspicuity of the microcalcifications. In particular, a difference image was formed of the results of application to the original image of a filter matched to a particular spot size and a median filter. An adaptive thresholding technique was used to label spots as calcifications, which spots also had to meet defined shape characteristics.
Accurate measurements of area, perimeter and shape of a spot are critical in classifying whether or not it corresponds to a microcalcification of a type associated with malignancy. Yet, prior art methods applied to make spots more conspicuous in radiologic images, such as that above-described, generally fail to preserve the locations of the edges or boundaries of the spots. Global spatial filtering with, for example, a Gaussian shaped filter, tends to smear these edges.
The problem of preservation of size and shape was addressed in J. Dengler et al. "Segmentation of Microcalcifications in Mammograms", (citation presently unknown). Therein, Gaussian high pass spatial filtering was used to remove the low frequency structural noise and "Difference of Gaussian" (DOG) spatial filtering was used to locate bright spots within a small range of sizes. Because this processing smoothed the boundaries of the spots, a morphological filter known as a "top-hat" transformation was applied to the original image to attempt to extract the boundaries of the spots. The spots located by the DoG filtering were iteratively expanded in size by topologically unimportant pixels, although not beyond the spot boundaries given by the morphological filter operation.
The efficacy of the method of J. Dengler et al. is dependent on proper choices for various spatial filter parameters (including positive and negative kernel widths, weight and threshold for the DoG filtering) which generally determine a range of sizes of spots that are detected as microcalcifications. Because of the irregular and frequently elongated nature of the microcalcification shapes of interest, it is not believed advisable to employ a size selective spatial filter which significantly restricts the range of spot sizes detected, particularly in a manner which is substantially independent of their shape. Further it does not seem wise to select candidate microcalcifications by thresholding intensity values. Such criteria may fail to select relatively dull spots which nonetheless have sharp edges.
Edge extraction is important in the general field of computer vision. An algorithm is discussed in P. Saint-Marc et al., "Adaptive Smoothing: A General Tool for Early Vision" IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 13, No. 6, pp. 514-529, 1991, which tends to maintain edge definition in filtered range images.
It is an object of the present invention to provide a method of aiding detection of cancer by utilizing computer detection of spots in digitized mammogram radiologic images in a manner which is not significantly dependent on the size of the spots and which preserves the locations of their edges or boundaries, from which edge locations, the spots are characterized by the computer as to whether they correspond to objects, in particular microcalcifications, of shape, edge contrast, brightness, and arrangement associated with malignancy. It is a further object of the present invention that such method involve a minimum number of parameters, which are where possible, computed from the image.
Briefly, these objects are satisfied by, with respect to a generated mammogram image stored in a digital storage device or medium accessible to a computer, computing filtered second spatial derivative of intensity values at the pixels of the image, in the form of a Laplacian, the zero crossings of which indicate edges. Regions of negative Laplacian value (which are therefore on the locally brighter side of the edges) are labelled and connected as locally bright spots. A plurality of feature measures are computed indicative of the respective shapes, edge contrasts and brightnesses of these locally bright spots, and a weighted combination of these feature measures is used to identify those spots corresponding to microcalcifications of a type associated with cancer. Thereafter, it is detected where the spots identified as microcalcifications are arranged in clusters, and the location of such clusters in the image are displayed on a monitor, or in some other human intelligible form.
The Laplacian values are determined in a manner that preserves the locations of edges. In particular, first spatial derivative of intensity values in the X and Y directions (i.e. gradient component values) are computed for each pixel position. Then, the first spatial derivative values in the X and Y directions are adaptive filtered by an iterative "adaptive smoothing" technique where in each iteration the strength or weakness of smoothing applied at each point in the image is determined from local conditions. In particular, measures of dispersion of differences between values of the first spatial derivative at adjoining pixel positions in the X and Y directions for the entire image are formed and smoothing applied to first spatial derivative values in the X and Y directions more strongly where there is little local variation in value relative to the respective dispersion measures, and more weakly where there is relatively great local variation. Values of the spatial derivative of the adaptively smoothed first derivative in X and Y directions are computed for each pixel position in the same respective directions to form second spatial derivative of intensity values in the X and Y directions. The Laplacian value is determined for each pixel position by adding the second spatial derivative values in the X and Y directions.
Other objects, features and advantages of the present invention will become apparent upon perusal of the following detailed description when taken in conjunction with the appended drawing, wherein:
FIG. 1 is a basic flow chart of the method of the present invention;
FIG. 2 is a flow chart of the detail of a filtered second derivative computation block in FIG. 1;
FIG. 3 is a flow chart of the detail of a connected negative region labelling block in FIG. 1;
FIGS. 4a through 4f are exemplary 2 D images displaying pixel values at various stages of image processing in accordance with the flow chart of FIG. 1, wherein, for the purpose of display, the intensities of the pixels of the images of FIGS. 4b through 4e are produced by adding a positive bias to their bipolar values equal to the largest negative value, so that a large negative value for a pixel appears as black; and
FIGS. 5a through 5f are exemplary graphs of pixel value versus position along a horizontal line 5--5 in FIG. 4a at the same stages of image processing as indicated in FIGS. 4a through 4f, respectively. The graph of FIG. 5f also appears in FIG. 5a as an aid in correlating between them.
Digitized images are readily available from imaging applications where X-ray image intensifier/camera chain or fluorescent image plate/laser readout or selenium plate/electrometer readout technologies are employed, without the necessity of scanning film. Such technologies are progressing in their spatial resolution and contrast sensitivities achieved and may soon find widespread use for mammographic applications. Presently, however, X-ray images of the human breast are generally obtained with film-screen technology. Based on studies by others of the resolution needed to properly represent individual microcalcifications in mammograms, available film scanners having spatial resolution of no worse than 0.1 mm per pixel and intensity amplitude resolution of at least 10 (and preferably 12) bits per pixel are required to produce a digitized image suitable for the purposes of the present invention. As shown in FIG. 1 of the drawing, the generation of the digitized X-ray image and its storage, for example as an image data file, on a suitable machine readable medium accessible to a computer workstation (not shown) such as a hard disk or RAM memory (not shown) forms the first step 10 in the method of the present invention. The computer workstation accesses the stored image information and performs various operations thereon, in the nature of image processing, in order to identify microcalcifications, and in particular to indicate clusters thereof. Such identification will aid early detection of breast cancer and could be used in a variety of operational modes as its efficacy is verified. Improving detection sufficiently to obviate the present practice of double reading would free significant radiologist reading resources.
As previously noted, the detection of microcalcifications of a type associated with malignancy requires significant experience. The important and most commonly noted characteristics of microcalcifications of this type are that they are relatively bright spots (in film-screen mammography) usually irregular in shape, e.g. elongated, and/or pointed, have sharp edge contrast with their surrounds, and most importantly, occur in clusters of at least 3 to 5 per square cm. in the radiologic film image. The irregular nature of these microcalcifications of interest indicate that it is crucial to preserve the location of edges in the identified spots so that suitable feature measures for classifying the spot can be accurately determined.
The Laplacian Of a Gaussian (LOG) is a widely published example of a filtered second spatial derivative operator. The technique of using the Zero Crossings of the Filtered Second Derivative (ZCFSD) for detecting and locating edges is based on a significant body of theoretical literature. The filtering, which is low pass, is designed to decrease the noise in the second derivative. Traditionally as in the LOG, the filter is Gaussian. As previously noted, such global spatial filtering has the disadvantage of moving edges away from their actual location in the image. I have employed a filtered second derivative which is produced in a manner that its zero crossings are located at edges in the image.
An overview of the present inventive method is apparent from the basic flow chart of FIG. 1. Therein, the original radiologic image generated and stored in box or step 10, which image comprises a rectangular array of pixels having respective intensities at each pixel position, is accessible for further processing. An example of a portion of an actual image is shown in FIG. 4a and of the intensities along a horizontal line therethrough in Figure 5a. The filtered second derivative thereof, in the form of a Laplacian, is computed in box 12, in a manner preserving the edge locations as zero crossings and connected regions of negative second derivative proximate the zero crossings are labelled by box 14. It may be recognized from examination of FIGS. 4e and 5e that regions of negative second derivative (which are black in FIG. 4e ) next to a zero crossing are on the bright side of an edge in the original image. The output of box 14 is a mask as shown in FIGS. 4f and 5f wherein locally bright regions or spots are white and the balance of the mask is black. Thereafter, in box 16, regions labelled by box 14 that are smaller than four pixels are deleted or removed from further consideration as candidates for microcalcifications of interest, by for example being set to black in the mask of locally bright regions. Similarly, labelled regions or bridges between portions of labelled regions which are only one pixel in width are deleted. Any regions split thereby are separately labelled. In box 18, feature measures are computed from the edge locations of the remaining locally bright spots. These feature measures are used in box 20 to identify or detect those of the locally bright spots which correspond to micro-calcifications of interest. Clusters containing at least three such microcalcifications in a square cm. are identified or detected in box 22. In box 24, these clusters are indicated by the computer in a human intelligible manner, e.g. on a CRT display (not shown) of the original image by circling, brightening or coloring the detected clusters on the display, thereby rendering them conspicuous. The procedures in the foregoing overview will now be explained in more detail.
The detail of the computation of the filtered second derivative is shown in FIG. 2. Therein, the first spatial derivatives of the original intensity image are computed in boxes 26 and 28 for the respective X and Y directions. The exemplary result of computation of the first spatial derivative in the X direction is shown in FIGS. 4b and 5b. These computations are made using the well known Sobel 3×3 gradient operators. Measures of the dispersions dx and dy of differences between first derivative values at adjoining pixel positions throughout the image in respective X and Y directions are determined in respective boxes 30 and 32. The first spatial derivatives in the X and Y directions are adaptive filtered in these directions by boxes 34 and 36, using an iterative adaptive smoothing, that adaptively smooths (in each iteration, convolves with a small averaging mask or kernel having locally adapted weights) first derivative values in the X and Y directions in a manner that there is relatively great smoothing applied where local variation in derivative values relative to the respective measures of dispersion dx and dy in these directions. In particular, a 3×3 averaging mask is used whose central relative weight to be applied to the first derivative value FDO at the considered pixel position is unity and whose other eight relative weights to be applied to first derivatives FDS at surrounding pixel positions touching the considered pixel position by a pixel side or corner is:
exp-((FDS -FDO /d)2
where d is respectively dx and dy for smoothing operations in the respective X and Y directions. I have found experimentally that, in blocks 30 and 32, computing the dispersions dx and dy as one half the range of the central third of the population of first derivative differences between adjoining pixel positions in the respective X and Y directions over the entire image, produces good results in combination with 9 iterations of application of the adaptive smoothing mask in boxes 34 and 36. The result of iterative adaptive smoothing the first spatial derivative of intensity in the X-direction for the exemplary image portion is shown in FIGS. 4c and 5c.
Spatial derivatives are computed in boxes 38 and 40 in respective X and Y directions of the adaptive filtered first spatial derivative of intensity values determined in boxes 34 and 36 to form filtered second derivative values in respective X and Y directions. FIGS. 4d and 5d illustrate the result of taking the spatial derivative in the X direction of the adaptive filtered first spatial derivative in the X direction. These spatial derivatives or gradients use the Sobel gradient operators. The filtered second spatial derivative values in the respective X and Y directions are added in box 42 to form a Laplacian value at each pixel position. The Laplacian is illustrated in FIGS. 4e and 5e for the exemplary image portion.
The boundaries of spots are determined from connected regions of negative Laplacian value. These are labelled in accordance with the flow chart of FIG. 3. In particular, in box 44, all pixel positions having negative Laplacian values are marked by a thresholding operation at the value zero. Those pixel positions having a Laplacian of less than zero are assigned a white value and those having non-negative Laplacian values are assigned a black value. In block 46, the interiors of marked connected regions of pixels (closed contours) are filled with white values. Each connected region is assigned a unique associated number in block 48. Block 16 then deletes regions smaller than four pixels and one pixel wide regions or bridges from further consideration, by for example, assigning the deleted regions or pixels to the black level. There results a mask of candidate locally bright regions or spots as illustrated in FIGS. 5f and 6f for the exemplary image portion.
For the purpose of discrimination of spots corresponding to mirocalcifications, I assume that microcalcifications are brighter than their surround, tend to have a compact shape, and they tend to have high edge gradients. It is true, however, that some microcalcifications of interest may appear to violate the assumption of compactness. However, this measure has been included to discriminate microcalcifications from scratches in the film and many fibrous objects such as ducts, blood vessels and support fibers.
Hence, in block 20, several measures are computed on the mask of locally bright regions or spots and their surrounding pixels which are indicative of the aforementioned characteristics or features, namely brightness (B), shape (S) and edge gradients (G). First, separate exterior and interior outline regions are grown as 4-connected neighborhoods around the boundary of the spot (i.e. pixels just within the spot and pixels just outside the spot that touch its boundary by a side of the pixel). The brightness measure (B) is computed as the average intensity of all pixels comprising the spot. To determine the shape measure S, there is first computed the area A of the spot, as the number of pixels comprising the spot, and the perimeter P of the spot, as the greater of the number of pixels comprising the exterior and interior outline regions. The shape measure S (which increases as a shape becomes less compact) is:
The gradient measure G is computed by 1) for each pixel in the exterior and interior outline regions adding the absolute values of the first derivatives in the X and Y directions determined in boxes 26 and 28, as an approximation of gradient magnitude; 2) computing respective averages of these approximate gradient magnitudes for the pixels of the exterior and interior outline regions; and 3) choosing as the gradient measure G, the greater of the averages for these two regions.
The feature measures are relatively weighted to give good results in classifying those spots for which the following inequality is true to be microcalcifications (where in is natural logarithm):
(9 ln B+7 ln G-2 ln S)>100
The detection in box 26 of whether the microcalcifications identified in box 22 are isolated or grouped in spatial clusters utilizes two techniques. First, data points identifying the locations of the centroids of the microcalcifications are clustered using the single link hierarchical method. All data points that are contained within an area of less than 1 cm square are grouped together. If the reader is unfamiliar with the mathematics of data clustering, see A. K. Jain et al., "Algorithms for Clustering Data", chapter 3, pp. 72-80, Prentice-Hall, Englewood Cliffs, N.J., 1988. Clusters with fewer than three members are considered to be benign, while clusters with 3 or more members are suspicious for malignant disease. It is the suspect clusters which are, by box 26, indicated in a display of the original image.
While the present invention has been described in particular detail in relation to identification of microcalcifications in mammographic images, it should also be appreciated that its principles have broad general applicability to the identification of spots in radiologic images which correspond to objects associated with malignancy, e.g. lung nodules in chest X-rays. Consequently, numerous modifications are possible in the details given which are within the intended spirit and scope. For example, the radiologic images might be the negative of those discussed with objects producing greater X-ray attenuation than their surround being locally dark.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US4515165 *||Sep 15, 1981||May 7, 1985||Energy Conversion Devices, Inc.||Apparatus and method for detecting tumors|
|US4618990 *||Nov 15, 1984||Oct 21, 1986||General Electric Company||Edge enhancement filtering for digital fluorography images|
|US4851984 *||Aug 3, 1987||Jul 25, 1989||University Of Chicago||Method and system for localization of inter-rib spaces and automated lung texture analysis in digital chest radiographs|
|US4907156 *||Jun 30, 1987||Mar 6, 1990||University Of Chicago||Method and system for enhancement and detection of abnormal anatomic regions in a digital image|
|US5003979 *||Feb 21, 1989||Apr 2, 1991||University Of Virginia||System and method for the noninvasive identification and display of breast lesions and the like|
|US5142557 *||Dec 21, 1990||Aug 25, 1992||Photometrics Ltd.||CCD and phosphor screen digital radiology apparatus and method for high resolution mammography|
|US5212637 *||Nov 22, 1989||May 18, 1993||Stereometrix Corporation||Method of investigating mammograms for masses and calcifications, and apparatus for practicing such method|
|FR2622714A1 *||Title not available|
|WO1991007135A1 *||Nov 20, 1990||May 30, 1991||Stereometrix Corporation||Method for assisting an investigation of a human breast for a malignancy and apparatus for practicing such method|
|1||"Image Processing for Recognition of Tumor on Mammography", S. Yabashi et al, Proceedings of the 1989 International Symposium on Noise and Cluster Rejection in Radars and Imaging Sensors, IEICE 1989.|
|2||Anil K. Jain, "Fundamentals of Digital Image Processing", Chapter 9, pp. 347-357, Prentice-Hall, Englewood Cliff, N.J., 1989.|
|3||Anil K. Jain, et al, "Algorithms for Clusterming Data", Chapter 3, pp. 72-80 Prentice Hall, Englewood Cliffs, N.J. 1988.|
|4||*||Anil K. Jain, et al, Algorithms for Clusterming Data , Chapter 3, pp. 72 80 Prentice Hall, Englewood Cliffs, N.J. 1988.|
|5||*||Anil K. Jain, Fundamentals of Digital Image Processing , Chapter 9, pp. 347 357, Prentice Hall, Englewood Cliff, N.J., 1989.|
|6||Atam P. Dhawan et al, "Mammographic Feature Enhancement by Computerized Image Processing", Computer Methods and Programs in Biomedicine, vol. 27, pp. 23-35, 1988.|
|7||*||Atam P. Dhawan et al, Mammographic Feature Enhancement by Computerized Image Processing , Computer Methods and Programs in Biomedicine, vol. 27, pp. 23 35, 1988.|
|8||Bahaa W. Fam et al, "Algorithm for the Detection of Fine Clustered Calcifications on Film Mammograms", Radiology, vol. 169, pp. 333-337, Nov. 1988.|
|9||*||Bahaa W. Fam et al, Algorithm for the Detection of Fine Clustered Calcifications on Film Mammograms , Radiology, vol. 169, pp. 333 337, Nov. 1988.|
|10||D. H. Davies et al, "Automatic Computer Detection of Clustered Calcifications in Digital Mammograms", Phys. Med. Biol. vol. 35, No. 8, 1112-1118, 1990.|
|11||D. H. Davies et al, "Automatic Detection of Microcalcifications in Digital Mammograms Using Local Area Thresholding Techniques", Medical Imaging III: Image Processing, vol. 1092, pp. 153-159, SPIE, 1989.|
|12||*||D. H. Davies et al, Automatic Computer Detection of Clustered Calcifications in Digital Mammograms , Phys. Med. Biol. vol. 35, No. 8, 1112 1118, 1990.|
|13||*||D. H. Davies et al, Automatic Detection of Microcalcifications in Digital Mammograms Using Local Area Thresholding Techniques , Medical Imaging III: Image Processing, vol. 1092, pp. 153 159, SPIE, 1989.|
|14||Hahaa W. Fam et al, "The Detection of Calcification Clusters in Film-Screen Mammograms; A Detailed Algorithmic Approach", Medical Imaging II, vol. 914, pp. 620-634, SPIE, 1988.|
|15||*||Hahaa W. Fam et al, The Detection of Calcification Clusters in Film Screen Mammograms; A Detailed Algorithmic Approach , Medical Imaging II, vol. 914, pp. 620 634, SPIE, 1988.|
|16||*||Heang Ping Chan et al, Computer Aided Detection of Microcalcifications in Mammograms Methodology and Preliminary Clinical Study , Investigative Radiology, vol. 23, pp. 664 671, Sep. 1988.|
|17||*||Heang Ping Chan et al, Image Feature Analysis and Computer Aided Diagnosis in Digital Radiography , Medical Physics, vol. 14, No. 4, pp. 538 548, Jul. 1987.|
|18||Heang-Ping Chan et al, "Computer Aided Detection of Microcalcifications in Mammograms-Methodology and Preliminary Clinical Study", Investigative Radiology, vol. 23, pp. 664-671, Sep. 1988.|
|19||Heang-Ping Chan et al, "Image Feature Analysis and Computer Aided Diagnosis in Digital Radiography", Medical Physics, vol. 14, No. 4, pp. 538-548, Jul. 1987.|
|20||*||Image Processing for Recognition of Tumor on Mammography , S. Yabashi et al, Proceedings of the 1989 International Symposium on Noise and Cluster Rejection in Radars and Imaging Sensors, IEICE 1989.|
|21||Isaac N. Bankman et al, "Automated Defection of Microcalcification Clusters in Mammograms", in S/CAR 90: Comp. App. to Assist Radiology, pp. 137-143, Symposia Foundation, Carlsbad, Calif. 1992.|
|22||*||Isaac N. Bankman et al, Automated Defection of Microcalcification Clusters in Mammograms , in S/CAR 90: Comp. App. to Assist Radiology, pp. 137 143, Symposia Foundation, Carlsbad, Calif. 1992.|
|23||Joachim Dengler et al, "Segmentation of Microcalcifications in Mammograms", submitted to IEEE Transactions on Medical Imaging, 1991.|
|24||*||Joachim Dengler et al, Segmentation of Microcalcifications in Mammograms , submitted to IEEE Transactions on Medical Imaging, 1991.|
|25||Laurens V. Ackerman et al, "Breast Lesion Classification by Computer and Xeroradiograph", Cancer, vol. 30, pp. 1025-1035, Oct. 1972.|
|26||*||Laurens V. Ackerman et al, Breast Lesion Classification by Computer and Xeroradiograph , Cancer, vol. 30, pp. 1025 1035, Oct. 1972.|
|27||Nico Karssemeijer, "A Stochastic Method for Automated Detection of Microcalcifications in Digital Mammograms", submitted to XIIth Conf. on Inf. Proc. in Med. Imaging, pp. 1-14, Jul. 1991.|
|28||*||Nico Karssemeijer, A Stochastic Method for Automated Detection of Microcalcifications in Digital Mammograms , submitted to XIIth Conf. on Inf. Proc. in Med. Imaging, pp. 1 14, Jul. 1991.|
|29||Petros Maragos et al, "Morphological Systems for Multidimensional Signal Processing", Proceedings of the IEEE, vol. 78, No. 4, Apr. 1990, pp. 690-710.|
|30||*||Petros Maragos et al, Morphological Systems for Multidimensional Signal Processing , Proceedings of the IEEE, vol. 78, No. 4, Apr. 1990, pp. 690 710.|
|31||*||Philippe Saint Marc et al, Adaptive Smoothing: A General Tool for Early Vision , IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 13, No. 6, Jun. 1991, pp. 514 529.|
|32||Philippe Saint-Marc et al, "Adaptive Smoothing: A General Tool for Early Vision", IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 13, No. 6, Jun. 1991, pp. 514-529.|
|33||Raj Jain, "The Art of Computer Systems Performance Analysis-Techniques for Experimental Design, Measurement, Simulation, and Modeling", pp. 192-197, John Wiley & Sons, New York, 1991.|
|34||*||Raj Jain, The Art of Computer Systems Performance Analysis Techniques for Experimental Design, Measurement, Simulation, and Modeling , pp. 192 197, John Wiley & Sons, New York, 1991.|
|35||Stanley H. Fox et al, "A Computer Analysis of Mammographic Microcalcifications: Global Approach", pp. 624-631, IEEE 1980, Proceedings of the 5th International Conf. on Pattern Recognition.|
|36||*||Stanley H. Fox et al, A Computer Analysis of Mammographic Microcalcifications: Global Approach , pp. 624 631, IEEE 1980, Proceedings of the 5th International Conf. on Pattern Recognition.|
|37||William G. Wee et al, "Evaluation of Mammographic Calcifications Using a Computer Programs", Radiology, vol. 116, pp. 717-720, Sep. 1975.|
|38||*||William G. Wee et al, Evaluation of Mammographic Calcifications Using a Computer Programs , Radiology, vol. 116, pp. 717 720, Sep. 1975.|
|39||Wolfgang Spiesberger, "Mammogram Inspection by Computer" IEEE Transactions on Biomedical Engineering, vol. BME 26, No. 4, Apr. 1979, pp. 213-219.|
|40||*||Wolfgang Spiesberger, Mammogram Inspection by Computer IEEE Transactions on Biomedical Engineering, vol. BME 26, No. 4, Apr. 1979, pp. 213 219.|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US5428655 *||Sep 10, 1993||Jun 27, 1995||Mitsui Mining & Smelting Co., Ltd.||Method and apparatus for three-dimensional detection of particles|
|US5583346 *||Jan 24, 1996||Dec 10, 1996||Fuji Photo Film Co., Ltd.||Method for detecting prospective abnormal patterns|
|US5586160 *||Mar 20, 1995||Dec 17, 1996||The Regents Of The University Of California||Automated analysis for microcalcifications in high resolution digital mammograms|
|US5615243 *||Mar 12, 1996||Mar 25, 1997||University Of Pittsburgh||Identification of suspicious mass regions in mammograms|
|US5740268 *||Sep 29, 1995||Apr 14, 1998||Arch Development Corporation||Computer-aided method for image feature analysis and diagnosis in mammography|
|US5757880 *||Jan 8, 1997||May 26, 1998||Colomb; Denis||Apparatus, article of manufacture, and method for creation of an uncompressed image of compressed matter|
|US5768333 *||Dec 2, 1996||Jun 16, 1998||Philips Electronics N.A. Corporation||Mass detection in digital radiologic images using a two stage classifier|
|US5825910 *||Oct 2, 1996||Oct 20, 1998||Philips Electronics North America Corp.||Automatic segmentation and skinline detection in digital mammograms|
|US5830141 *||Sep 27, 1996||Nov 3, 1998||U.S. Philips Corporation||Image processing method and device for automatic detection of regions of a predetermined type of cancer in an intensity image|
|US5857030 *||Apr 9, 1996||Jan 5, 1999||Eastman Kodak Company||Automated method and system for digital image processing of radiologic images utilizing artificial neural networks|
|US5933518 *||Apr 22, 1996||Aug 3, 1999||U.S. Philips Corporation||Method and device for image processing for automatically detecting objects in digitized images|
|US5987094 *||Oct 30, 1997||Nov 16, 1999||University Of South Florida||Computer-assisted method and apparatus for the detection of lung nodules|
|US5994713 *||Jul 25, 1997||Nov 30, 1999||Quantum Imaging Corp.||Filmless photon imaging apparatus|
|US6091841 *||Oct 14, 1999||Jul 18, 2000||Qualia Computing, Inc.||Method and system for segmenting desired regions in digital mammograms|
|US6115488 *||Oct 14, 1999||Sep 5, 2000||Qualia Computing, Inc.||Method and system for combining automated detections from digital mammograms with observed detections of a human interpreter|
|US6137898 *||Nov 25, 1998||Oct 24, 2000||Qualia Computing, Inc.||Gabor filtering for improved microcalcification detection in digital mammograms|
|US6167146 *||Oct 14, 1999||Dec 26, 2000||Qualia Computing, Inc.||Method and system for segmentation and detection of microcalcifications from digital mammograms|
|US6173034 *||Jan 25, 1999||Jan 9, 2001||Advanced Optical Technologies, Inc.||Method for improved breast x-ray imaging|
|US6205236||Oct 12, 1999||Mar 20, 2001||Qualia Computing, Inc.||Method and system for automated detection of clustered microcalcifications from digital mammograms|
|US6246782||Jun 6, 1997||Jun 12, 2001||Lockheed Martin Corporation||System for automated detection of cancerous masses in mammograms|
|US6389157||Jan 11, 2001||May 14, 2002||Qualia Computing, Inc.||Joint optimization of parameters for the detection of clustered microcalcifications in digital mammograms|
|US6477262 *||Jun 26, 2001||Nov 5, 2002||Shih-Ping Wang||Computer-aided diagnosis method and system|
|US6556699||Aug 24, 2001||Apr 29, 2003||Qualia Computing, Inc.||Method for combining automated detections from medical images with observed detections of a human interpreter|
|US6574357||Sep 5, 2001||Jun 3, 2003||Shih-Ping Wang||Computer-aided diagnosis method and system|
|US6650766||Oct 25, 2002||Nov 18, 2003||Qualia Computing, Inc.||Method for combining automated detections from medical images with observed detections of a human interpreter|
|US6694046 *||Mar 28, 2001||Feb 17, 2004||Arch Development Corporation||Automated computerized scheme for distinction between benign and malignant solitary pulmonary nodules on chest images|
|US6757415||Jun 13, 2003||Jun 29, 2004||Qualia Computing, Inc.||Method for determining features from detections in a digital image using a bauer-fisher ratio|
|US6763128||Jun 13, 2003||Jul 13, 2004||Qualia Computing, Inc.||Method for analyzing detections in a set of digital images using case based normalcy classification|
|US6801645||Jun 23, 2000||Oct 5, 2004||Icad, Inc.||Computer aided detection of masses and clustered microcalcifications with single and multiple input image context classification strategies|
|US6970587||Sep 29, 2003||Nov 29, 2005||Icad, Inc.||Use of computer-aided detection system outputs in clinical practice|
|US7103224 *||Jan 20, 2004||Sep 5, 2006||Virtualscopics, Llc||Method and system for automatic identification and quantification of abnormal anatomical structures in medical images|
|US7106893 *||Nov 18, 2003||Sep 12, 2006||The Johns Hopkins University||Method and apparatus for segmenting small structures in images|
|US7231071 *||Sep 12, 2002||Jun 12, 2007||Fujifilm Corporation||Abnormal shadow detecting system|
|US7272250 *||Nov 23, 2001||Sep 18, 2007||R2 Technology, Inc.||Vessel segmentation with nodule detection|
|US7308126||Oct 14, 2005||Dec 11, 2007||Icad, Inc.||Use of computer-aided detection system outputs in clinical practice|
|US7383237||Feb 6, 2006||Jun 3, 2008||Health Discovery Corporation||Computer-aided image analysis|
|US7397937||Sep 30, 2002||Jul 8, 2008||R2 Technology, Inc.||Region growing in anatomical images|
|US7430309||Feb 8, 2005||Sep 30, 2008||Institut De Cardiologie De Montreal||Computation of a geometric parameter of a cardiac chamber from a cardiac tomography data set|
|US7502524 *||Aug 20, 2003||Mar 10, 2009||Nxp B.V.||Method and apparatus of processing a skin print image|
|US7769244 *||Mar 4, 2008||Aug 3, 2010||Algosoft-Tech Usa, Llc.||Automatic digital film and video restoration|
|US8086002 *||Apr 9, 2008||Dec 27, 2011||Three Palm Software||Algorithms for selecting mass density candidates from digital mammograms|
|US8892188 *||Feb 3, 2006||Nov 18, 2014||Koninklijke Philips N.V.||Identifying abnormal tissue in images of computed tomography|
|US9089307||Oct 21, 2010||Jul 28, 2015||Koninklijke Philips N.V.||Three-dimensional analysis of lesions represented by image data|
|US20030081822 *||Sep 12, 2002||May 1, 2003||Fuji Photo Film Co., Ltd.||Abnormal shadow detecting system|
|US20030099386 *||Sep 30, 2002||May 29, 2003||Schneider Alexander C.||Region growing in anatomical images|
|US20040109592 *||Nov 18, 2003||Jun 10, 2004||Bankman Isaac N.||Method and apparatus for segmenting small structures in images|
|US20040114829 *||Oct 10, 2003||Jun 17, 2004||Intelligent System Solutions Corp.||Method and system for detecting and correcting defects in a digital image|
|US20040146204 *||Jan 20, 2004||Jul 29, 2004||Virtualscopics, Llc||Method and system for automatic identification and quantification of abnormal anatomical structures in medical images|
|US20050201598 *||Feb 8, 2005||Sep 15, 2005||Francois Harel||Computation of a geometric parameter of a cardiac chamber from a cardiac tomography data set|
|US20050232472 *||Aug 20, 2003||Oct 20, 2005||Koninklijke Philips Electronics N.V.||Method and apparatus of processing a skin print image|
|US20060171573 *||Oct 14, 2005||Aug 3, 2006||Rogers Steven K||Use of computer-aided detection system outputs in clinical practice|
|US20080170801 *||Mar 4, 2008||Jul 17, 2008||Algosoft Limited||Automatic digital film and video restoration|
|US20080267470 *||Apr 9, 2008||Oct 30, 2008||Three Palm Software||Algorithms for selecting mass density candidates from digital mammograms|
|US20080269598 *||Feb 3, 2006||Oct 30, 2008||Koninklijke Philips Electronics N.V.||Identifying Abnormal Tissue in Images of Computed Tomography|
|US20100150413 *||Jun 23, 2006||Jun 17, 2010||Konica Minolta Medical & Graphic, Inc.||Abnormal shadow candidate detecting method and abnormal shadow candidate detecting device|
|US20120318998 *||Feb 15, 2011||Dec 20, 2012||Vladimir Kondrasovs||On-line measurement method for ionizing radiation|
|USRE43152||Sep 12, 2008||Jan 31, 2012||The Johns Hopkins University||Method and apparatus for segmenting small structures in images|
|USRE43894||Dec 7, 2011||Jan 1, 2013||The Johns Hopkins University||Method and apparatus for segmenting small structures in images|
|WO1996021199A1 *||Nov 29, 1995||Jul 11, 1996||Philips Electronics N.V.||Mass detection by computer using digital mammograms of the same breast taken from different viewing directions|
|WO1999028853A1 *||Nov 25, 1998||Jun 10, 1999||Qualia Computing, Inc.||Automated detection of clustered microcalcifications from digital mammograms|
|WO2012085818A1||Dec 19, 2011||Jun 28, 2012||Koninklijke Philips Electronics N.V.||Mammography calcium score|
|International Classification||G06F19/00, G06T7/00|
|Cooperative Classification||G06F19/321, G06T2207/30068, G06F19/3487, G06T7/0012|
|European Classification||G06F19/34P, G06T7/00B2|
|Jan 11, 1993||AS||Assignment|
Owner name: NORTH AMERICAN PHILIPS CORP., NEW YORK
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST.;ASSIGNOR:CARMAN, CHARLES SPENCER;REEL/FRAME:006400/0939
Effective date: 19930107
|Apr 29, 1998||FPAY||Fee payment|
Year of fee payment: 4
|Jun 4, 2002||REMI||Maintenance fee reminder mailed|
|Nov 15, 2002||LAPS||Lapse for failure to pay maintenance fees|
|Jan 14, 2003||FP||Expired due to failure to pay maintenance fee|
Effective date: 20021115